GPT-6.1 Terra Puts the Cost of Successful Computer Work in Focus
OpenAI's new model lowers standard token prices, making completion rates and correction costs central to the purchasing decision.
Published October 27, 2026Briefs / Artificial Intelligence / Generative Models
OpenAI released GPT-6.1 Terra on October 26 for visual reasoning, computer use and writing. Its changelog lists availability in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise and Edu plans, subject to enterprise workspace and administrator controls. Developers can access it through the application programming interface (API) as gpt-6.1-terra. Standard API prices are $1 per million input tokens, $0.05 for cached input and $5 for output. Prompts above 272,000 input tokens cost more; cache writes and alternate processing modes have separate rates. [1] [2] [3] [4]
OpenAI reports gains on its computer-work evaluations, including a 15-percentage-point improvement over GPT-5.6 Terra on Agents' Last Exam at medium reasoning effort. Those vendor results suggest a model worth testing on repetitive desktop tasks. A buyer's relevant measure is spending per accepted completion: model and tool charges across every attempt, plus supervision and correction, divided by work that meets the required standard. Cheap attempts can still become expensive when failures consume staff time. [1]
A representative trial should include awkward forms, ambiguous documents and recovery from mistakes, then record accepted results, retries and human effort. The model documentation says tool calling requires the Responses API, an integration detail that also shapes the test. Evidence from that full workflow would show where Terra's lower token prices translate into a lower cost of getting useful work done. [2]